Two pieces published this week on LinkedIn deserve the attention of investment management professionals. Ray Dalio argues with characteristic bluntness that we are not witnessing a collection of isolated regional crises — we are in the early stages of a world war that, in his words, "isn't going to end anytime soon." Mohamed El-Erian provides the market-level translation: the IMF is now warning that regardless of how the Middle East conflict evolves, "all roads lead to higher prices and lower growth."
Taken together, these two perspectives paint a picture that should give every quantitative investor serious pause: on portfolio positioning, and about the very tools they are using to support decision-making, especially their stock-selection models.
Dalio's framework describes something investment professionals should take seriously: the possibility of a genuine regime change in the world order. He outlines a thirteen-step cycle of escalating conflict and economic disorder, and argues that we are currently at Step 9 — multi-theatre conflicts happening simultaneously. He draws explicit parallels to the 1913–14 and 1938–39 periods, moments that were, by definition, outside the distribution of any model trained on peacetime data.
El-Erian reinforces this from a market perspective. He describes a world where parts of Asia and Africa are already moving into what he calls Phase Three of economic damage — not just higher energy costs, but genuine demand destruction driven by fears over physical energy availability. Europe is weeks behind. The US, while relatively better positioned, is seeing manufacturing price pressures hit their highest level since 2022, with the ISM Prices Paid index jumping from 70.5 to 78.3. This is the type of environment in which traditional, static, backwards-looking models begin to fail.
Factor models built on decades of return history assume a degree of structural stability in correlations, volatility regimes, and cross-asset relationships. When geopolitical stress fractures those relationships — as it is doing now, with El-Erian noting that the usual correlation between US equities and oil prices broke down last week — those models do not “just” underperform, they can actively mislead.
Back-tested strategies are, by construction, trained on a world that no longer exists in the same form. When the regime changes, the backtest becomes archaeology.
This is where Axyon AI's approach becomes directly relevant to the scenarios described.
1. Early adaptation to Emerging Themes
Regime shifts do not announce themselves; they unfold through underlying themes that progressively reshape the market gradually in shifting narratives. In the current environment, the ability to identify and construct thematic universes around emerging opportunities before they become consensus trades is a material and defensible source of alpha. This is where agentic AI can be a great ally in identifying themes and translating them into investment universes.
2. Adaptive learning in non-stationary environments
Axyon AI's machine learning models are designed to continuously update their understanding of market dynamics rather than rely on fixed, historically calibrated parameters. Following a robust process, the models identify persistent patterns across markets and generate predictive AI-based signals about the future relative performance of securities.
3. Signal detection beyond traditional factors
The current Middle East situation is now generating cascading effects across energy markets, inflation data, central bank policy, and G7 political cohesion simultaneously. Traditional factor models, which tend to decompose returns along established dimensions such as value, momentum, or quality, were not designed to capture this kind of multi-dimensional, event-driven stress propagation. AI-driven models can identify non-linear patterns and emergent signals across a far broader feature space.
Dalio observes that in prolonged conflicts, the winner is rarely the most powerful side — it is the side that can endure the most pain the longest. A similar dynamic sometimes plays out in investment management. There is an oft-cited quotation worth remembering here:
"The market can stay irrational longer than you can stay solvent."
A manager can be fundamentally right about a position, a factor exposure, or an investment process, and still be out of business before the market comes around. Static models, by their nature, may put managers in exactly this situation — defending an out-of-favour process, absorbing drawdowns, and hoping the regime reverts before clients begin to lose confidence. AI offers a fundamentally different posture: rather than enduring until an uncertain vindication, the model actively seeks new information to adapt your process in real time.
When traditional factor models are under stress, when static backtests are anchored to a world that has shifted, and when correlations that held for years are breaking down in real time, the question is:
Is your model learning, or is it remembering?
Axyon AI provides institutional-grade Predictive and Agentic AI-driven solutions to support investment management decisions. To learn more about how our models are designed for non-stationary market environments, please get in touch.